The present invention relates in general to communication systems, and is particularly directed to a new and improved modulation scheme that is especially suited for QPSK-based satellite communication systems. The modulation scheme is effective to rob a relatively limited portion of the available transmitter power, and inject into the QPSK waveform a prescribed amount of carrier energy, through which detection and recovery of the carrier at the receiver may be achieved without incurring a signal to noise degradation penalty. In addition, the injected carrier-based modulation scheme of the invention exploits the substantially improved performance of modern forward error correction codes, and significantly reduces the signal power required for achieving a relatively low bit error rate.
A major concern of both providers and users of satellite communication systems is how to maximize the use of system resources. The most important resources are considered to be transponder bandwidth and effective isotropic radiated power (EIRP), since some portion of each is employed by every signal sent through the transponder. Because satellite resources are expensive (for example, a single transponder may cost hundreds of thousands of dollars per month in leasing fees), for the case where satellite power is the scare resource, minimizing the amount of power required for each signal allows more signals to be sent through the transponder, and thereby reduces leasing fees. An alternative application is to reduce the aperture size of the receiver antenna for the same transponder power. More recently developed low-cost systems that use small aperture antennas tend to be power-limited as they have lower G/T values, and therefore require more power from the satellite.
FIGS. 1 and 2 are examples of simulated spectrum analyzer displays of satellite transponder utilization. In the spectral utilization of FIG. 1, wherein the available passband is essentially xe2x80x98packedxe2x80x99 with signals, optimum utilization is obtained when all of the transponder""s available EIRP is employed. The utilization diagram of FIG. 2, on the other hand, shows the case where more than half of the transponder""s bandwidth goes unused. This condition indicates that money is wasted, since users are paying for the entire transponder, but utilizing -only a portion of its available capacity. Such power-limited utilization of the transponder may be due to the receiving antennas on the ground having relatively small apertures, so that more power is required for adequate signal quality.
Earth terminals of commercial satellite communication systems have historically employed relatively large, and therefore large gain-to-noise temperature (G/T) ratio, antennas. Since these systems tend to be bandwidth-limited, considerable effort has gone into developing more bandwidth-efficient modulation techniques, such as using some form of M-ary phase shift keying (MPSK) and quadrature amplitude modulation (QAM). Much less work has been carried out in improving power efficiency than in improving bandwidth efficiency. If more power-efficient modulation techniques were available, then each signal would require less power, and a larger number signals could be sent through a power-limited transponder. Alternatively, if the amount of power a given signal requires can be minimized, the required earth terminal EIRP and hence transmitter and/or antenna aperture size can be minimized. This is a third major benefit to small-aperture systems, which enjoy: 1- reduced satellite power usage; 2- reduced transmitter power or antenna aperture for the ground terminal; and 3- reduced antenna aperture for the receive terminal. The first and second benefits go together, while the third may be considered a trade-off against the first and second.
FIG. 3 diagrammatically illustrates the modulation and demodulation signal schemes employed by respective transmitting and receiving earth stations 10 and 20 that are linked by a satellite transponder 30 of a typical QPSK system. Historically, QPSK (and also BPSK) has been a preferred modulation scheme for satellite communications since, among other advantages, no additional energy is required to transmit a discrete carrier reference. Instead, the demodulator is responsible for restoring or xe2x80x98regeneratingxe2x80x99 the carrier based on the received signal.
At the transmit site 10, quadrature channel data symbols dI and dQ, that have been encoded with some form of forward error correction (FEC) code, are modulated in mixers 11I and 11Q onto respective phase-quadrature components of a carrier signal fC. As will be discussed in detail below, the use of forward error correction encoding of the data serves to trade bandwidth for power. The phase quadrature modulated signals are then summed in a summer 13 into a composite QPSK signal. This QPSK signal, a spectral waveform for which is shown in FIG. 4, is transmitted via amplifier-feed circuitry 14 coupled to an antenna 15.
At the receive site 20, signals received by an antenna 22 and associated low noise amplifier circuitry 23 are coupled to a demodulator loop, which supplies both I and Q carrier references. To demodulate the data, the received signal is coupled to a carrier recovery or regeneration path 25 and a data recovery path 27. As shown in the spectral diagram of FIG. 4, since no discrete carrier component is separately transmitted from the transmit site 10, the carrier must be xe2x80x98regeneratedxe2x80x99 at the receive site 20.
For QPSK signals this is usually accomplished by means of a relatively complex circuit 26, such as a Costas loop, or a fourth-power circuit, so as to provide a carrier reference. Its output drives a phase locked loop 28, so as to provide a carrier reference for the data recovery path. The data recovery path 27 includes a phase detector 29I/Q, to which the received I/Q channel data plus carrier and the regenerated carrier signals are supplied. The output of the phase detector 29I/Q represents the encoded data symbols, which are applied to downstream error correction recovery circuitry to recover the original data.
As described above with reference to the transponder utilization diagram FIG. 2, a large percentage of transponder bandwidth often goes unused, so that improving power efficiency will allow more signals to be transmitted through the same transponder. In fact, using more bandwidth to gain power efficiency is a good trade in many systems, as there will still be sufficient bandwidth to support additional users. One way to trade bandwidth for power is to avoid the use of modulation waveforms, such as QAM, that give up power efficiency for bandwidth. As shown in system diagram of FIG. 3, described above, another technique is to use forward error correcting codes. In addition to the use of FEC codes, error detection and retransmission can be used to minimize transponder power usage.
Forward error correcting codes trade bandwidth for power by sending redundant symbols in order to enable errors to be corrected at the receive site. Forward error correction has a long history in satellite communication systems and many types of decoders are available as inexpensive chips. Some codes employ check bits to verify that no errors were made in the reception. If an error is detected, then the receiving site requests that the transmitter site re-send the block of data where the error appeared. This can be a difficult technique for communication over geosynchronous satellites, due to the long time delays involved. Protocols have been developed with these delays in mind, and many systems now employ both error detection and retransmission. Still, in heavy fading conditions, as can occur during rainstorms, the system may often become clogged with retransmissions. As a result, performing all error correction at the receiver is highly desirable, even if retransmission is used.
At present, the most commonly used error correcting codes are convolutional codes, typically running at rate xc2xd, wherein two coded symbols are transmitted for every one information symbol, thus doubling the transmission rate and hence the occupied bandwidth. One way to gain efficiency at the expense of bandwidth is to use even lower-rate codes, such as rate ⅓ codes. Another common technique is to concatenate two codes. This most often takes the form of concatenating a convolutional code with a block code, such as a Reed-Solomon code. These two types of codes have good synergy, and significant power gains can be realized with relatively little additional band-spreading.
A significant problem with these types of codes is that they do not necessarily work well at a very low energy per bit-to-noise density ratio (Eb/N0)xe2x80x94on the order of one to zero dB. While these codes are capable of yielding ultra-low error rates at moderate Eb/N0 values, they do not produce a significant drop in required power for moderate error rates (their efficiency falls off rapidly below about 4 dB Eb/N0). Commercially available demodulators are built with this limitation in mind, and do not provide carrier tracking below about 4 dB Eb/N0.
Demodulation with a low-rate code, such as rate xc2xd or rate ⅓, at a very low Eb/N0 is difficult for two reasons: first- carrier phase and symbol timing are very difficult to recover; second- maintaining soft decision thresholds is also a problem. A high ratio of the symbol rate to the data rate implies a very low ES/N0. For example, for a rate ⅓ code, ES/N0 is about 5 dB less than Eb/N0. The demodulator must have sufficient bandwidth to pass these high symbol rates, and thus must work at an extremely low signal to noise ratio (S/N). As discussed above, carrier recovery generally involves a nonlinear operation, such as raising the signal to a power (e.g., fourth power in the carrier regenerator circuit 26 of FIG. 3). When carried out at a very low signal to noise ratio, the signal to noise ratio is reduced even further, making the demodulator""s task of recovering the data extremely difficult.
FIG. 5 demonstrates the carrier recovery problem using a fourth power circuit for QPSK signals. (Other means of carrier regeneration such as Costas loops offer approximately the same performance as the fourth power device.) To track down to 0 dB ES/N0 using a rate ⅓ code for QPSK demodulation, the input SNR is about xe2x88x922 dB and the output SNR is about xe2x88x9238 dB. This causes a very difficult, if not effectively impossible acquisition problem, which mandates the use of extremely narrow tracking loops to maintain an accurate phase estimate.
In accordance with the present invention, the above discussed power and bandwidth utilization problems associated with conventional carrierless PSK modulation schemes employed for satellite communication systems are effectively obviated by means of a modulation waveform that is effective to rob a relatively limited portion of the available transmitter power, and inject into the QPSK waveform a prescribed amount of carrier energy, for example at about 15 dB below the signal. This injected carrier component serves to facilitate detection and recovery (rather than non-linear regeneration) of the carrier at the receiver.
In addition, the injected carrier-based modulation scheme of the invention exploits the substantially improved performance of modern forward error correction coding schemes, such as but not limited to turbo codes, so as to significantly reduce the signal power required for achieving a relatively low bit error rate. Namely, using a prescribed xe2x80x98carrier-injectingxe2x80x99 waveform in combination with advanced coding techniques enables the invention to double the number of small-terminal users on a satellite transponder, thus effectively halving their leasing costs. Conversely, the waveform of the invention allows use of a smaller antenna for more flexible siting or mobility.
Pursuant to a first embodiment, a DC offset or bias is applied to the encoded data stream upstream of the QPSK modulation circuitry, which serves to shift reference levels for the encoded in-phase data symbol stream to values that cause the spectral waveform of the QPSK signal to contain a prescribed amount of carrier signal energy. This readily discernible injected carrier obviates the need for a non-linear carrier regeneration circuit in the carrier recovery path of the receiver, as in a conventional QPSK demodulator. Instead, the carrier may be directly extracted by a carrier recovery phase locked loop.
Although transmitting a carrier at about 15 dB below the signal results in very little performance reduction, the bias loss in QPSK causes degradation of the signal to noise ratio. For xe2x88x9215 dB in carrier leakage, the loss is about 0.3 dB at low ES/N0 values, and becomes progressively worse at higher signal levels. It would be very desirable to avoid this loss; in addition, the transmission of an unmodulated carrier component may result in violation of spectral density requirements, such as international communication standards that limit the energy density that may be irradiated upon the earth from a satellite transponder.
To satisfy these objectives a second embodiment of the invention injects into the QPSK waveform a xe2x80x98spreadxe2x80x99 carrier, that is functionally equivalent to that achieved by the direct insertion of a DC bias voltage in the first embodiment. Since the loss incurred with QPSK is caused by a difference in level between a xe2x80x981xe2x80x99 and a xe2x80x980xe2x80x99, the difference can be made to average to zero over a symbol time by chopping the transmitted carrier with a pseudo random square wave having values +1 and xe2x88x921. If the carrier is spread at the symbol rate, with an edge occurring at mid-bit, for example, the signal level will be averaged within the demodulator matched filter and no degradation will result. This restores the QPSK loss to be the same as for BPSK.
If the carrier is spread using a square wave (i.e., an alternating 1xe2x88x920 pattern, the transitions are aligned so as to occur at the midpoint of the data symbols. This ensures that the carrier bias averages to zero over a data symbol time. However, if pseudo-random spreading is used, transitions do not occur at every symbol. This permits some data symbols to be biased. By using Manchester (bi-phase) coding on the spreading sequence, a transition is guaranteed at each symbol and the bias is removed.
To recover the carrier, the demodulator of the second embodiment uses a despreading sequence generator identical to that used in the modulator since the spreading operation is synchronous with symbol timing. The received signal is bandpass-filtered and down-converted to a complex baseband signal. To extract the carrier, the complex baseband signal is coupled to a despreading mixer along with a despreading PN waveform supplied by a phase locked loop tuned to the symbol rate. The extracted carrier is filtered using a phase lock loop which provides a coherent carrier reference to each of a pair of in-phase and quadrature channel mixers, to which the complex baseband signal is applied. The outputs of the mixers are filtered in matched filters, symbol timing for which is derived by filtering the data component of the chopped spectrum using a phase lock loop. The recovered data samples and the symbol clock are coupled to a downstream decoder.
In addition to injecting a prescribed amount of carrier energy into the transmitted QPSK waveform, the present invention takes advantage of more powerful modern codes, such as but not limited to xe2x80x98turboxe2x80x99 codes, which are capable of extending moderate error rate performance down to low values of Eb/N0. High performance codes tend to be directed more toward achieving ultra-low bit error rates at moderate Eb/N0 values, rather than moderate bit error rates at ultra-low Eb/N0, so that there is no real savings in transmitter power in using commercial modems, which are generally designed to hold in no lower than about 4 dB Eb/N0. To realize a sufficient transmitter power savings, the modem must be capable of recovering a coherent carrier reference at negative ES/N0, which cannot be achieved by a conventional demodulator.